|
Processing a subset of a picture
Rick Richardson from United States [1 posts] |
13 year
|
So I have figured out how to spot a laser point, but I need to caculate the center of a pixel down to the subpixel range.
I also need to limit search area of the picture down to a few vertical lines, i.e., I know the laser will be in the middle 15 verticle lines of the pictures because of a fixed relationtionship between the video sensor and the laser. I have figured out a way to do it but I thought I would ask you the most effecient way to search that restricted area. The best solution is to read only that area from the camera, but that is a driver issue.
Does anyone know if there is code for a driver for the Logitech Orbit that is open source?
|
|
|
|
Anonymous |
13 year
|
Rick,
You will have to interpolate the pixel values in order to determine what subpixel point is the highest point. See
http://www.tinaja.com/glib/pixintpl.pdf
for some examples.
Unless you want to get into some very low level programming I think the best you can do for cropping is to do so after you get the image from the camera. In your case you would just skip the first X pixels and XX last pixels. You may also check to see what image size gets you the least horizontal pixels ...
This wouldn't be for a 3D laser scanner would it? Have you seen the one on the Neato vacuum robot? Lots of interest in that lately.
We're not aware of any open source driver for the Orbit.
STeven.
|
|
|
|
Rick from United States [4 posts] |
13 year
|
Thanks for the info. As for the Neato sensor, I actually bought your software before I knew about the Neato. I had read an article about triangulation laser scanning and have been working on it. The Neato IEEE article refines some of the math and highlights some issues.
The Neato is a good solutions for a small robot that wants to avoid a few things, but is not a great one for a larger robot. I believe the time for a cheaper laser scanner is here and companies just don't know it. I have started a prototype using PVC, a stepper motor and a lazy susan bearing ring. That's where most of my questions come from.
Right now, I am trying to figure out how to find a cheap slip ring that can last at 600 rmp and source a cmos linear array something with 1200x100 and global shutter or some combination better that the big sensors I am stuck with. I probably don't need more than 600x25, but all the ones I can get my hands on are 512x1.
Any ideas or thoughts would be appreciated.
What would be cool is using your software to collect the data points and then putting them on top of the realtime picture and adjusting the data points based on the changing realtime picture.
|
|
|
|
Anonymous |
13 year
|
Rick,
We don't have much to offer in terms of sensor advice but are certainly interested in assimilating the data for reconstruction after the device is built. We agree that cheaper laser scanners are in great demand and would be a welcome product to most robotic engineers. That's in part seen by the enthusiasm for devices like the Kinect and Neato products that may not be of the quality of a traditional LIDAR system but are certainly much less expensive.
Let us know how it goes!
STeven.
|
|
This Thread